Skip to content

System_Architecture.md

Latest commit

 

History

History
87 lines (68 loc) · 4.62 KB

System_Architecture.md

File metadata and controls

87 lines (68 loc) · 4.62 KB

System Architecture: Self-Balancing Robot

This document outlines the core structural design and hardware-software interaction layers governing the Self-Balancing Robot platform. The architecture is explicitly decoupled, isolating rigid mathematics from variable external inputs to mathematically guarantee chassis stability.

High-Level Abstraction

The processing pipeline is segmented functionally into three primary execution domains:

  1. Sensory & Fusion Layer: Retrieves and cleans raw external environmental data.
  2. Kinetic Control Layer: The mathematical core translating state disparity into reactionary acceleration vectors.
  3. Actuation & Hardware Layer: Responsible strictly for generating perfectly uniform discrete electrical outputs.
graph TD
    classDef hardware fill:#0b1d30,stroke:#2a5275,stroke-width:2px,color:#ffffff
    classDef software fill:#16322b,stroke:#2a5f51,stroke-width:2px,color:#ffffff
    classDef network fill:#312015,stroke:#704020,stroke-width:2px,color:#ffffff

    subgraph Sensory Layer
        IMU[ISM6HG256X IMU]:::hardware -->|I2C 400kHz| FUSION[Mahony AHRS Fusion]:::software
        MAG[QMC5883L Compass]:::hardware -->|I2C 400kHz| FUSION
        FUSION -->|Cascaded EMA Filter| STATE[Robot State: Pitch, Yaw]:::software
    end

    subgraph External Inputs
        JOY[Joystick Transmitter]:::network -->|ESP-NOW| RECV[ESPNOW Receiver]:::software
        TUNE[Bluetooth Serial Tuner]:::network -->|UART| UART_T[Param Parser]:::software
        RECV --> TGT[Target Angle / Steering]:::software
        UART_T --> C[Tuning Configurations]:::software
    end

    subgraph Kinetic Control
        STATE --> PID_BAL[Balance PID Loop]:::software
        STATE --> PID_YAW[Heading PID Loop]:::software
        TGT --> PID_BAL
        TGT --> PID_YAW
        C --> PID_BAL
        PID_BAL --> MIX[Kinematic Mixer]:::software
        PID_YAW --> MIX
    end

    subgraph Actuation Layer
        MIX -->|Steps/sec| TIMER[Timer ISR Bresenham 20kHz]:::software
        TIMER -->|Fast UART| TMC1[Left TMC2208]:::hardware
        TIMER -->|Fast UART| TMC2[Right TMC2208]:::hardware
        TMC1 --> M1[Left NEMA17 Stepper]:::hardware
        TMC2 --> M2[Right NEMA17 Stepper]:::hardware
    end
Loading

1. Core Component Interactions

Sensory and Fusion (I2C at 400kHz)

The rigid foundational measurement stems from an ISM6HG256X high-performance 6-axis IMU, supplemented by a QMC5883L magnetometer for absolute heading tracking.

  • Raw values enter the Mahony AHRS mathematical filter.
  • The 9-DOF quaternion natively generated shields pitch measurements from magnetic interference, rendering a structurally pure 3D map of the chassis orientation.
  • Secondary cascaded Exponential Moving Average (EMA) algorithms surgically scrub remaining electromagnetic noise out of the finalized pitch angle variable before it ever reaches the control loops.

The Mathematics Engine (200 Hz Main Loop)

A rigidly constrained microcontroller loop evaluates exactly 200 times per second ($dt = 0.005s$).

  1. Identifies external translation requests (from Bluetooth or ESP-NOW Joystick).
  2. The core Balance PID measures the discrepancy between mathematical upright and actual pitch.
  3. The Heading PID isolates any lateral variance restricting vehicle travel direction.
  4. Calculations are aggregated through a limiter capping acceleration change rates.

Precision Stepper ISR (20 kHz Hardware Timer)

This layer operates completely detached from the primary logical sequence.

  • The physics engine feeds absolute "Steps Per Second" integers exclusively into volatile memory registers.
  • A background Interrupt Service Routine triggers exactly 20,000 times per second, managing Bresenham Accumulators for each wheel.
  • When an accumulator boundary is violated, native silicon overrides trigger the electrical step pin immediately, achieving zero-jitter mechanical pulses.

2. Communications Ecosystem

The ESP32 processes two disparate concurrent wireless network environments:

ESP-NOW Remote Manipulation

  • An ultra-fast, connectionless protocol fundamentally identical to raw WiFi routing.
  • The transmitter establishes fixed payload vectors representing intent, not mechanical realities.
  • The robot dynamically incorporates this intent as mathematically filtered offsets against its own gravitational setpoint.

Bluetooth Classic Serial Diagnostics

  • An asynchronous text parser operating on BluetoothSerial, permitting complex string evaluations.
  • Translates dynamic textual configuration instructions directly into variable manipulation modifying PID structural constraints natively on the fly without demanding code recompilation.